Scribd
Upload
256 views

Measurement Errors

Specification & measurement errors in linear regression models can occur through the inclusion of irrelevant variables, omission of relevant variables, and measurement errors in variables. Multicollinearity is a case where independent variables are highly correlated, and can be caused by data collection methods, constraints on the population or model, or an overdefined model. Autocorrelation in time series data occurs when observations are not independent, which can result from omitted variables, model misspecification, or measurement errors. Heteroscedasticity violates the assumption of constant error variance across independent variables, which can arise from grouped rather than individual data or in time series data.

Uploaded by

Diana Moturi
Copyright
© © All Rights Reserved
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
256 views

Measurement Errors

Specification & measurement errors in linear regression models can occur through the inclusion of irrelevant variables, omission of relevant variables, and measurement errors in variables. Multicollinearity is a case where independent variables are highly correlated, and can be caused by data collection methods, constraints on the population or model, or an overdefined model. Autocorrelation in time series data occurs when observations are not independent, which can result from omitted variables, model misspecification, or measurement errors. Heteroscedasticity violates the assumption of constant error variance across independent variables, which can arise from grouped rather than individual data or in time series data.

Uploaded by

Diana Moturi
Copyright
© © All Rights Reserved
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 2

Specification & Measurement Errors

These types of errors occur when developing a linear regression model. A linear regression
model will include choosing the independent variables and defining the omitted variables as well
as choosing the functional form. Specification errors will therefore mainly occur in the following
ways:

Inclusion of irrelevant variables the inclusion of an irrelevant variable does not have an
impact on the consistency of the effects of the variables on the model. However, the
irrelevant variables will generally be inefficient due to the fact that their variances will be

greater than those that exist in the model.


Omission of relevant variables - This occurs when in determining which variables will be
omitted from the linear regression model. The effect of omitting a relevant variable will
lead to the generation of bias and inconsistent results when estimating the effects of the

independent variables on the dependent one.


Measurement errors in variables this type of error assumes that both the dependent and
independent variables are accurate. Measurement error therefore occurs when we cannot
accurately measure the magnitude of the variable of interest. This can occur either in the
measurement of dependent or independent variable.

Multicollinearity
This is a case of multiple regression in which the independent variables are highly correlated. It
measures how two variables affect each other and to what extent. It is the degree of correlation
and not whether it is present or not.
Multicollinearity is caused by the following

The data collection method employed - The data collection method can lead to
multicollinearity problems when the analyst samples only a subspace of the region of the

independent variables that are defined in an equation.


Constraints on the model or in the population if there is a constraint on the population

sample then there will always be multicollinearity.


Model specification if the range of a model is increased then the chance of
multicollinearity occurring also increase significantly.

An over defined model these types of models generally tend to have more independent
variables than the number of observations.

Autocorrelation
Autocorrelation occurs when the observations of the dependent variable are not independently
drawn. This normally occurs in the case of time series data. Autocorrelation doesnt occur in
cross sectional data where the individual units are independent of each other. Usually an error is
carried over from one period to another.
Autocorrelation can be caused by the following factors:

It may be caused by the omission of various variables.


It may also be caused by misspecification. This is caused by the misinterpretation of an

element within the regression model.


The third reason may be caused by the systematic errors in measurement. This usually
occurs when an error is inputted when creating the regression model.

Heteroscedasticity
This occurs when applying the standard regression model we assume that the variance of the
error term remains constant, conditional on the independent variables. This assumption fails
when the error term changes in the different variables within the regression model. This for
example occurs if the variance of the error term changes when there is a change in any of the
independent factors within the regression model, when this occurs then the error process is
heteroskedastic.
Heteroscedasticity can occur in the following ways:

It may arise also when one uses grouped data rather than individual data.
It can occur in time series data also.

You might also like